AI

Authsnap wins AI pitch contest by taking on rural hospitals’ denial workload

Six founders faced questions about adoption, clinical escalation and who pays for AI. The winning pitch paired appeal-letter automation with the work of fitting small hospitals’ workflows.

Author
HDM Staff
Save
Share
Authsnap’s Gretchen Heinen with the AI Startup Pitch Competition emcee and judges
Authsnap’s Gretchen Heinen joins the emcee and judges after the AI Startup Pitch Competition. Photo credit: HLTH USA 2026

    Authsnap won a unanimous decision in HLTH USA 2026’s AI Startup Pitch Competition with a proposition aimed at hospitals that cannot afford a large technology implementation: turn clinical records and an insurance denial into an appeal letter, while helping understaffed teams do the work around it. Founder and CEO Gretchen Heinen argued that the shortage of people to prepare clinically grounded appeals makes rural providers a different market from large hospital systems.

    The November 17 Revv Stage contest also exposed a common hurdle across six very different businesses. Producing an AI output was only part of their challenge; each needed a credible way to get data, fit an existing workflow, involve clinicians when necessary and persuade an organization to buy.

    Google for Startups’ Samantha Perrone served as emcee. Judges were Google for Startups’ Jen Diffenderfer and David White, and NVIDIA’s Chelsea Sumner, a clinical pharmacist who leads the company’s healthcare AI startup ecosystem. They questioned founders about implementation, commercial differentiation, accessibility, safety and return on investment before selecting Authsnap.

    Appeals need a workflow as well as a letter

    Heinen, a nurse, described Authsnap’s approach as a comparatively light implementation for rural hospitals and clinics. Staff can supply clinical information and a denial reason in a PDF, with the software preparing an appeal letter. She said the company takes a share tied to successful recovery, rather than requiring a small provider to undertake a lengthy enterprise integration before starting.

    “We don't have enough nurses to write these letters in rural communities or at clinics,” Heinen said. She described organizations where one person, sometimes a physician working after hours, carries that burden. Her pitch emphasized the practical difference between making letter generation available and giving a thinly staffed organization the capacity to use it consistently.

    Asked about the biggest obstacle, Heinen pointed to the surrounding work: “It's really getting in there and reworking their workflows.” That can require services that do not scale like software alone, she acknowledged. She said nurses still handle much of chart intake, while a self-service model and further automation remain development goals.

    Heinen reported three hospital customers and several distribution relationships since commercialization in July. Those were company-reported commercial developments, not independently verified evidence of appeal success or cash recovered. She also explained why the company moved from serving individual patients toward business customers: the emotional support involved in individual appeals did not fit the model it believed could have the largest financial impact.

    A swallowing screen starts with one clinical population

    Amanda “Mandy” Mahoney, CEO and co-founder of Accelowave Technologies, pitched the Aspirometer, a sensor-based approach to swallowing assessment. Her proposed initial market was certified stroke centers, where screening is already part of the workflow, rather than every potential clinical setting at once.

    “It's a clinical support tool to help nurses make faster and more accurate decisions,” Mahoney said. She described a neck sensor, a handheld device and algorithms intended to return a pass-or-refer result. The company’s development description places clinical interpretation and regulatory review on its roadmap; the pitch should not be read as an announcement of an approved, fully deployed diagnostic product.

    Sumner’s question about children brought an important boundary into view. “They have different anatomy,” Mahoney said, explaining that the work had been in adults and pediatric applications would require additional data. Mahoney also said medical-record connectivity remained in development and that the company was raising money to reach a clinical trial. Those answers made the proposed expansion distinct from the evidence available for the initial population.

    Real-world evidence depends on access to real-world data

    Firas Rhaiem, CEO and co-founder of Clarrio.ai, described combining clinical, treatment, laboratory and other patient information into longitudinal records. His interest began with trying to understand his sister’s epilepsy, he said. The business seeks patterns linking disease, medication and behavior outside the limited time patients spend with a clinician.

    “It takes ages to accomplish things,” Rhaiem said when asked about barriers. He identified interoperability as a major problem, with different health records and regulatory environments complicating deployment across sites. He reported work with 10,000 patients and pharmaceutical proof-of-concept projects, but the presentation did not provide a study design sufficient to establish that its reported retention figures would generalize to another trial or population.

    “We spent two years doing POCs,” he said, describing the time needed to build the relationships and data connections behind the platform. His answer to the competition question was partly about that accumulated implementation work. A rival might reproduce technology, he argued, while still needing time to reproduce the connections and experience.

    AI governance has to reach the deployed agent

    Will Haver, founder and CEO of Skovos, pitched an approach to governing hospital AI agents, including mapping tools, controlling access and investigating failures. The company’s leadership page identifies Skovos as a trade name of Haver AI, the name used in earlier coverage. Haver’s presentation focused on making governance practical after an agent is already running.

    “We govern AI agents for hospitals and health systems,” Haver said. He described an initial mapping and sandbox offering, followed by annual licensing per agent. His reported revenue and health-system adoption were commercial claims from the pitch, rather than an audited assessment presented to the audience.

    Asked how the service wins a place in constrained hospital budgets, Haver emphasized the governance committee. “This governance committee is composed not just of the CIO,” he said, also identifying AI, compliance and legal leadership. His answer illustrated the number of decision-makers a safety product must satisfy; showing a technical control is different from proving that it reduces incidents or earns a sustainable budget.

    The difficult part of follow-up is escalation

    Royi Rozen, founder of MedComm, described an AI care orchestrator that calls and texts patients after emergency care, reinforces instructions, helps arrange follow-up and alerts clinical teams to problems. He said pilot work and health-record integrations were underway, with additional patients to be enrolled in the coming weeks. That timing matters: planned enrollment is not a completed outcomes study.

    “We are closing the follow up care gap,” Rozen said. Questions then moved from the appeal of automated outreach to the circumstances under which it would fail, including patients who cannot be reached and symptoms requiring a person’s intervention.

    “We're building those clinical escalation layers,” he said, describing work with clinical teams to distinguish issues for a nurse from emergencies. He also said enrolling patients before discharge helps them recognize the service’s calls. The pitch made clear why generating a voice interaction can be quicker than defining, testing and integrating the clinical actions that must follow it.

    Ovarian-age information needs a route into care

    Kiran Kumar, CEO of Timeless Biotech, presented MenoTime, which uses existing blood-work data to estimate ovarian age and time to menopause. The company describes it as clinical decision support, and Kumar positioned distribution through physicians as central to the business. She did not propose that an AI prediction alone should replace a clinician’s assessment.

    “We did decide to go purely through physicians,” Kumar said when asked about direct-to-consumer access. Her reasoning was that information is hard to translate into treatment without a physician and that many women seek help through routine care rather than a separate concierge service.

    “With payers, it's very big on cost avoidance,” Kumar said of the commercial argument. She linked that proposition to reducing unnecessary assessments during the menopause transition. The pitch supplied a proposed mechanism for savings, rather than a controlled comparison showing that the product itself had delivered those savings across a payer population.

    The judges’ final decision favored Authsnap’s broad addressable need, with Diffenderfer announcing the unanimous selection. Across the contest, the strongest practical question remained the same: whether a promising output can become a reliable service inside healthcare’s actual constraints. For the winner, that meant recognizing that rural hospitals need help with the work around an appeal, not simply a faster way to produce its text.